EP1404152A2 - Device and method for fitting a hearing-aid - Google Patents

Abstract

The adaption method uses provision of evaluation data for different defined hearing situations and adaption of the hearing aid via individual weightings provided for the hearing aid wearer, with a continuous weighting function passing through support points obtained by individual weighting of the evaluation data corresponding to the actual hearing situation. Also included are Independent claims for the following: (a) a device for adaption of a hearing aid to different hearing situations for the individual hearing aid wearer; (b) a hearing aid adapted to different hearing situations for the individual hearing aid wearer

Description

The present invention relates to a method for adapting a hearing device by providing evaluation data for different, predetermined listening situations and adapting the hearing device to a hearing device wearer by means of individual weighting. In addition, the present invention relates to a corresponding device for adjusting a hearing aid and an individually adjustable hearing aid.

From the document DE 690 12 582 T1 a hearing aid is known, which the user can set menu-driven individually. By lightly touching a control pad, the user gains access to a new parameter set for a specific response function, which is then input to a digital signal processor. By means of a few touches, the user can find the answering function appropriate to his acoustic environment and the required gain. Furthermore, from the document US 4,731,850 a programmable digital hearing aid system is known. By programming, the electro-acoustic properties of the hearing aid can be adapted to the patient and the environment. Selected parameter values are loaded into programmable memory (EEPROM) which provides corresponding coefficients to a programmable filter and amplitude limiter of the hearing aid so as to achieve automatic adjustment for ambient noise, speech levels and the like.

Basically, there is the danger for the hearing device wearer that the hearing device confuses a detected hearing situation. If such a confusion occurs, the hearing aid adjusts itself with his hearing aid parameters to a hearing situation that is currently not available. Thus, the audio signals are transmitted to the hearing aid wearers inappropriate. For example, if the listening situation "language at rest" with the Hearing situation "music" confused, so under certain circumstances unnecessary or disturbing frequency components are transmitted or certain frequency components improperly amplified.

In current hearing aids there is often an unclear relationship between a specially detected hearing situation and the hearing aid parameters. In many cases, the relationship between detected hearing situations and corresponding hearing device settings in the current state of the art is also very simply realized. In noise situations, for example, the directional microphone and the noise reduction is activated. A classifier recognizes and classifies a current listening situation and toggles between a selection of hearing aid programs with a variety of parameters. However, there is the problem that a current hearing situation does not readily correspond to a standardized, typical hearing situation. Accordingly, there is some uncertainty as to which hearing device program the hearing device should switch or which hearing aid parameters would have to be set for the optimal use of the hearing device. Typical problem cases are mixed situations, when, for example, speech is to be transmitted against the background of music and other background noises.

The object of the present invention is therefore to improve the adaptation of a hearing device to a current hearing situation.

According to the invention, this object is achieved by a method for adapting a hearing device by providing evaluation data for various predetermined hearing situations and fitting the hearing device to a hearing device wearer by means of individual weighting, wherein the individual weighting is performed by a continuous weighting function passing through interpolation points, each of which is an individual Weighting of the evaluation data represent one of the given listening situations.

Furthermore, the above-mentioned object is achieved according to the invention by a device for adapting a hearing device with a memory device for providing evaluation data for various predefined listening situations and an adaptation device for adapting the hearing device to a hearing aid wearer by means of individual weighting, wherein the individual weighting by means of a continuous weighting with the adaptation device Weighting function is executed, which passes through vertices, which each represent an individual weighting of the evaluation data of one of the predetermined listening situations of the memory device.

In an advantageous manner, the hearing device parameters can thus be continuously adapted to different listening situations. The abrupt change of a complete hearing device parameter set can thereby be avoided, so that a current auditory situation does not have to be discretely assigned to a predetermined class.

Conveniently, the evaluation data are determined offline in advance by a sound signal analysis. Thus, a database having multiple evaluation data for a plurality of listening situations can be constructed as bases for a continuous function. The evaluation data may include weight vectors with respect to specific audio signals that are characteristic of the given listening situations. Such weight vectors can advantageously be determined by an eigenvector analysis of the specific audio signals.

In a so-called fitting analysis, the weighting function for the individual weighting can be determined from listening situations characteristic of the hearing aid wearer. Thus, it is possible to deal specifically with the habits of the hearing aid wearer and those hearing situations that occur during the hearing him most frequently used as a basis for the adjustment of the hearing aid.

The weighting function is conveniently determined from at least one fitting parameter and at least one value of the evaluation data. To refine the individualization of a hearing device, it is also possible to use a plurality of values of the evaluation data for obtaining the weighting function.

Figur 1

ein Ablaufdiagramm für eine Offline-Schallsignalanalyse;

Figur 2

ein Ablaufdiagramm für eine Offline-Anpassanalyse; und

Figur 3

ein Ablaufdiagramm für eine Echtzeit-Klassifikation.

The present invention will now be explained in more detail with reference to the accompanying drawings, in which:

FIG. 1

a flow chart for an offline sound signal analysis;

FIG. 2

a flow chart for an offline fitting analysis; and

FIG. 3

a flow chart for a real-time classification.

The exemplary embodiments described below represent preferred embodiments of the present invention. According to the invention, the method for adapting a hearing device to a hearing device wearer or his hearing loss includes two offline methods and one real-time method. First, in an off-line sound signal analysis, a plurality of typical audio signals are analyzed for characteristic evaluation data. Subsequently, in an offline fitting analysis, a fitting function that is individual for a hearing device wearer with the characteristic evaluation data is obtained as a parameter. Finally, in a real-time method with the help of the obtained fitting function, the hearing aid is adjusted individually for a current hearing situation.

Specifically, offline sound signal analysis is used to determine generic listening situations that make up hearing situations how to put together "language in peace" or "music" or mix them together. The advantage of considering generic listening situations is that they can be clearly separated. Mathematically, these generic listening situations are described by feature vectors that are orthogonal to each other and result from a principle component analysis (PCA) of the feature vectors of common listening situations. However, common listening situations, such as music, speech, etc., are not orthogonal to each other and thus can not be clearly separated from each other. The description of common listening situations by generic listening situations in the form of orthogonal feature vectors reduces the further data processing effort enormously. The results of a PCA are essential input for further steps.

In the course diagram of FIG. 1, the essential steps of an offline sound signal analysis are shown in principle. In a step 10, first N-classes of listening situations are determined. Such classes would be for example: H ₁ = speech at rest, H ₂ = loud speech at rest, H ₃ = speech at noise, H ₄ = music etc.

In step 11, M signal characteristics that may be altered by the digital signal processing of the hearing aid are defined. Such signal characteristics would be, for example: F _{1... I} = spectral envelope (LPC coefficients), F _{i... J} = modulation power density spectrum, etc.

In a subsequent step 12, Q-typical audio signals are collected for each listening situation {x _i } _Hj . These then correspond to a sound sample database for the different listening situations.

Then, according to step 13, the characteristics of the audio signals collected in step 12 are determined. These result in F _ijk = F _i ({X _j } _Hk ), i = 1..M, j = 1..Q, k = 1..N.

For each hearing situation, the feature correlation is determined individually (a) and total (b) in step 14. This results in the correlation matrices C _a and C _b .

Finally, in step 15, the eigenvectors corresponding to the generic listening situations and the intrinsic features of the correlation matrices C _a and C _{b are} determined by diagonalizing. Furthermore, the normalized eigenvalues (statistical weights) are determined for the subsequent fitting process.

In this context, for example, the speech feature vector V _max and generic feature vectors V _{gi are} determined. The speech feature vector V _max corresponds to the C _a eigenvector for "speech at rest" having the highest eigenvalue. By contrast, the generic feature vectors V _gi represent the n C _b eigenvectors with the highest eigenvalues with which, for example, 95% of all audio signals can be restored.

The feature vector of any audio signal may be considered as a superposition of generic feature vectors: F = a ₁ * V _g1 + a ₂ * V _g1 + ... where a ₁ , ..., a _{n denote} the weighting vectors of a specific audio signal.

The probability that any audio signal corresponds to the typical listening situation "speech at rest" is: $${\text{p = F * V}}_{\text{Max}}$$

With the offline sound signal analysis are thus determined by correlation of the individual features, such as modulation depth, modulation frequency, energy in a frequency band, etc., the main features or main eigenvectors typical listening situations. The weights of the main features represent, as already mentioned, about 95% of the sum of all weights, so that the other features are negligible.

Each typical hearing situation can thus be relatively clearly characterized by a few key features.

The offline fitting analysis serves on the one hand to determine an individual base adjustment, z. As the hearing aid fitting, which judges a particular hard of hearing person as optimal for speech at rest. On the other hand, the offline fitting analysis is used to determine the required parameter changes depending on the mixing ratio of the generic listening situations. The result is a functional relationship between the mixing parameters of a given hearing situation and the individual and optimal hearing aid parameters for this situation. The advantage of this is that the hearing aid parameters that are suitable for a hearing situation are determined individually for the hearing aid wearer and can be changed fluently in fluent transitions of hearing situations, since the functional relationship was determined. This procedure should be implemented in the hearing aid fitting software because the function that maps the mixing parameters to the hearing aid parameters must be determined with the fitting software and programmed into the hearing aid.

The individual hearing loss of a patient is taken into account in the offline fitting or fitting analysis as follows. The patient is first interviewed in step 20 for characteristic hearing situations in his social environment. Here he calls those listening situations that have the most importance for him or occur most frequently such as "speech at rest", "phone calls", etc.

For this purpose, several suitable audio examples are selected from the audio databases created according to steps 10 to 12. The record x ₀ corresponds to z. For example, the audio sample "Language at rest". There are n different audio examples x ₀ ... x _n available.

In step 22, the weight vectors a _0... A _{n of} the selected sound examples are determined. They are taken from the database created during the offline sound signal analysis.

The best individual fit with corresponding match parameter vectors is determined according to step 23. For this purpose, for example, the approach of interactive, adaptive fitting is selected for the sound examples. The corresponding fitting or fitting parameter vectors are b ₀ ... b _n . This step ensures a subjective assessment of typical, objective listening situations.

Finally, in step 24, a function is determined with which individual adjustments due to changes in the weight vectors can be carried out continuously. For example, using the values a ₀ and b ₀ as a reference, it is possible to predict individual fitting changes as a function of the weighting changes. The complexity of this prediction or its accuracy depends on the dimension of the vectors a and b, ie the number of analyzed features and the number of fitting parameters. The result is a function b = b ₀ + φ (| a ₀ -a |) or b = b ₀ + c ₁ | a ₀ -a | + c ₂ | a ₀ -a | ² + ... The Taylor coefficients c ₁ , c ₂ ... can be determined by regression. The determined function, based on one or more coefficients, thus quantizes the relationship between objective hearing situation and subjective perception.

The real-time classification or real-time setting of the hearing device makes it possible that upon detection of a specific mixing ratio of generic hearing situations, the corresponding hearing device parameter set is active and the transitions are fluid.

The individual function determined in steps 20 to 24 in FIG. 2 is used during the operation of the hearing device for real-time classification according to the method sequence of FIG. 3 used. In this real-time setting of the hearing aid, according to step 30, a main adjustment parameter is used for basic adjustment of the hearing aid. The main adjustment parameter b ₀ classifies the hearing situation which is individually most important for the patient.

In step 31, the feature vector of the input signal is determined as a function of time F = F (x). The basis of this determination is the input signal in a time window, with which the feature vector results just for this window.

The weighting vector is determined in step 32 according to the above-described function F = a ₁ * V _g1 + a ₂ * V _g1 + ..., as a function of time.

With the aid of the individual fitting function b = b ₀ + φ (| a ₀ + a |) determined in step 24, the best individual setting or adaptation of the hearing device to the current hearing situation is performed in step 33. It is possible to consider mixed situations continuously and to adjust the hearing aid to the individual needs of the patient or hearing aid wearer.

For this purpose, the adjustment vector is finally smoothed in step 34.

The advantage of this real-time classification is the relatively low computational cost of M multiplies, where M equals the number of features. In addition, relatively small memory space is required, namely M bytes. However, about N additional control signals are required, where N corresponds to the number of controlled hearing aid parameters.

According to the invention, an individualization with regard to the setting of a hearing device as well as an improved adaptation to mixtures of typical hearing situations is thus possible.

The device according to the invention or the method according to the invention greatly reduces confusion between detected hearing situations. There is a clear mapping of listening situations to hearing aid parameters as well as an individual classification.

Claims (14)

Method for adjusting a hearing aid
Providing rating data (steps 10 to 15) for various predetermined listening situations and
Fitting the hearing aid (steps 20 to 34) to a hearing aid wearer by means of individual weighting,
characterized in that
the individual weighting (step 24) is performed by a continuous weighting function passing through vertices each representing an individual weighting of the scoring data of one of the predetermined auditory situations. The method of claim 1, wherein the evaluation data is determined by a sound signal analysis (steps 10 to 15). The method of claim 1 or 2, wherein the evaluation data comprises weight vectors relating to specific audio signals characteristic of the given listening situations. The method of claim 3, wherein the weight vectors are determined by eigenvector analysis (step 15) of the specific audio signals. Method according to one of claims 1 to 4, wherein the weighting function for the individual weighting is determined from hearing aid wearer characteristic hearing (step 20). Method according to one of claims 1 to 5, wherein the weighting function is determined from at least one fitting parameter and at least one value of the evaluation data. Method for operating a hearing device by recording an audio signal of a current hearing situation,
Calculating signal evaluation data from the audio signal (step 31),
Weighting the signal evaluation data (steps 32 and 33) by means of a continuous weighting function obtained according to one of claims 1 to 6, and
Adapting the hearing device according to the weighted signal evaluation data to the current hearing situation, in particular under real-time conditions. Device for adjusting a hearing aid with
a memory device for providing evaluation data for various predetermined listening situations and
an adaptation device for adapting the hearing aid to a hearing aid wearer by means of individual weighting,
characterized in that
the individual weighting can be carried out with the adaptation device by means of a continuous weighting function which runs through interpolation points, which in each case individually weight the evaluation data of one of the predefined auditory situations from the memory device represent. Apparatus according to claim 8, comprising a sound signal analysis device with which the evaluation data for the predetermined situations can be determined and from which the evaluation data can be transferred to the memory device. Apparatus according to claim 8 or 9, wherein the evaluation data comprises weight vectors relating to specific audio signals characteristic of the given listening situations. Apparatus according to claim 10, comprising an analysis device with which the weight vectors can be determined by eigenvector analysis of the specific audio signals. Device according to one of claims 8 to 11, which has an offline setting device for determining the weighting function for the individual weighting from hearing aid wearer characteristic hearing situations. Apparatus according to claim 12, wherein the weighting function can be determined from at least one fitting parameter and a plurality of the evaluation data by the offline setting device. Hearing aid with
a recording device for recording an audio signal of a current listening situation,
a calculator for calculating signal evaluation data from the audio signal,
a weighting means for weighting the signal evaluation data by means of a continuous weighting function and
a control or regulating device for adapting the hearing device according to the weighted signal evaluation data to the current hearing situation, in particular under real-time conditions.

Images